1/*-
2 * Copyright (c) 2011 NetApp, Inc.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD$
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD$");
31
32#include <sys/types.h>
33#include <sys/mman.h>
34#include <sys/time.h>
35
36#include <machine/segments.h>
37
38#include <stdio.h>
39#include <stdlib.h>
40#include <string.h>
41#include <err.h>
42#include <libgen.h>
43#include <unistd.h>
44#include <assert.h>
45#include <errno.h>
46#include <pthread.h>
47#include <pthread_np.h>
48#include <sysexits.h>
49
50#include <machine/vmm.h>
51#include <vmmapi.h>
52
53#include "bhyverun.h"
54#include "acpi.h"
55#include "inout.h"
56#include "dbgport.h"
57#include "legacy_irq.h"
58#include "mem.h"
59#include "mevent.h"
60#include "mptbl.h"
61#include "pci_emul.h"
62#include "pci_lpc.h"
63#include "xmsr.h"
64#include "ioapic.h"
65#include "spinup_ap.h"
66#include "rtc.h"
67
68#define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
69
70#define	VMEXIT_SWITCH		0	/* force vcpu switch in mux mode */
71#define	VMEXIT_CONTINUE		1	/* continue from next instruction */
72#define	VMEXIT_RESTART		2	/* restart current instruction */
73#define	VMEXIT_ABORT		3	/* abort the vm run loop */
74#define	VMEXIT_RESET		4	/* guest machine has reset */
75
76#define MB		(1024UL * 1024)
77#define GB		(1024UL * MB)
78
79typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
80
81char *vmname;
82
83int guest_ncpus;
84
85static int pincpu = -1;
86static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
87static int virtio_msix = 1;
88
89static int foundcpus;
90
91static int strictio;
92
93static int acpi;
94
95static char *progname;
96static const int BSP = 0;
97
98static int cpumask;
99
100static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
101
102struct vm_exit vmexit[VM_MAXCPU];
103
104struct bhyvestats {
105        uint64_t        vmexit_bogus;
106        uint64_t        vmexit_bogus_switch;
107        uint64_t        vmexit_hlt;
108        uint64_t        vmexit_pause;
109        uint64_t        vmexit_mtrap;
110        uint64_t        vmexit_inst_emul;
111        uint64_t        cpu_switch_rotate;
112        uint64_t        cpu_switch_direct;
113        int             io_reset;
114} stats;
115
116struct mt_vmm_info {
117	pthread_t	mt_thr;
118	struct vmctx	*mt_ctx;
119	int		mt_vcpu;
120} mt_vmm_info[VM_MAXCPU];
121
122static void
123usage(int code)
124{
125
126        fprintf(stderr,
127                "Usage: %s [-aehAHIPW] [-g <gdb port>] [-s <pci>] [-S <pci>]\n"
128		"       %*s [-c vcpus] [-p pincpu] [-m mem] [-l <lpc>] <vm>\n"
129		"       -a: local apic is in XAPIC mode (default is X2APIC)\n"
130		"       -A: create an ACPI table\n"
131		"       -g: gdb port\n"
132		"       -c: # cpus (default 1)\n"
133		"       -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
134		"       -H: vmexit from the guest on hlt\n"
135		"       -P: vmexit from the guest on pause\n"
136		"       -W: force virtio to use single-vector MSI\n"
137		"       -e: exit on unhandled I/O access\n"
138		"       -h: help\n"
139		"       -s: <slot,driver,configinfo> PCI slot config\n"
140		"       -S: <slot,driver,configinfo> legacy PCI slot config\n"
141		"       -l: LPC device configuration\n"
142		"       -m: memory size in MB\n",
143		progname, (int)strlen(progname), "");
144
145	exit(code);
146}
147
148void *
149paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
150{
151
152	return (vm_map_gpa(ctx, gaddr, len));
153}
154
155int
156fbsdrun_disable_x2apic(void)
157{
158
159	return (disable_x2apic);
160}
161
162int
163fbsdrun_vmexit_on_pause(void)
164{
165
166	return (guest_vmexit_on_pause);
167}
168
169int
170fbsdrun_vmexit_on_hlt(void)
171{
172
173	return (guest_vmexit_on_hlt);
174}
175
176int
177fbsdrun_virtio_msix(void)
178{
179
180	return (virtio_msix);
181}
182
183static void *
184fbsdrun_start_thread(void *param)
185{
186	char tname[MAXCOMLEN + 1];
187	struct mt_vmm_info *mtp;
188	int vcpu;
189
190	mtp = param;
191	vcpu = mtp->mt_vcpu;
192
193	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
194	pthread_set_name_np(mtp->mt_thr, tname);
195
196	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
197
198	/* not reached */
199	exit(1);
200	return (NULL);
201}
202
203void
204fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip)
205{
206	int error;
207
208	if (cpumask & (1 << vcpu)) {
209		fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
210		    vcpu);
211		exit(1);
212	}
213
214	cpumask |= 1 << vcpu;
215	foundcpus++;
216
217	/*
218	 * Set up the vmexit struct to allow execution to start
219	 * at the given RIP
220	 */
221	vmexit[vcpu].rip = rip;
222	vmexit[vcpu].inst_length = 0;
223
224	mt_vmm_info[vcpu].mt_ctx = ctx;
225	mt_vmm_info[vcpu].mt_vcpu = vcpu;
226
227	error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
228	    fbsdrun_start_thread, &mt_vmm_info[vcpu]);
229	assert(error == 0);
230}
231
232static int
233vmexit_catch_reset(void)
234{
235        stats.io_reset++;
236        return (VMEXIT_RESET);
237}
238
239static int
240vmexit_catch_inout(void)
241{
242	return (VMEXIT_ABORT);
243}
244
245static int
246vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
247		     uint32_t eax)
248{
249#if BHYVE_DEBUG
250	/*
251	 * put guest-driven debug here
252	 */
253#endif
254        return (VMEXIT_CONTINUE);
255}
256
257static int
258vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
259{
260	int error;
261	int bytes, port, in, out;
262	uint32_t eax;
263	int vcpu;
264
265	vcpu = *pvcpu;
266
267	port = vme->u.inout.port;
268	bytes = vme->u.inout.bytes;
269	eax = vme->u.inout.eax;
270	in = vme->u.inout.in;
271	out = !in;
272
273	/* We don't deal with these */
274	if (vme->u.inout.string || vme->u.inout.rep)
275		return (VMEXIT_ABORT);
276
277	/* Special case of guest reset */
278	if (out && port == 0x64 && (uint8_t)eax == 0xFE)
279		return (vmexit_catch_reset());
280
281        /* Extra-special case of host notifications */
282        if (out && port == GUEST_NIO_PORT)
283                return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
284
285	error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
286	if (error == 0 && in)
287		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
288
289	if (error == 0)
290		return (VMEXIT_CONTINUE);
291	else {
292		fprintf(stderr, "Unhandled %s%c 0x%04x\n",
293			in ? "in" : "out",
294			bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
295		return (vmexit_catch_inout());
296	}
297}
298
299static int
300vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
301{
302	fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
303	    *pvcpu);
304	return (VMEXIT_ABORT);
305}
306
307static int
308vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
309{
310	int newcpu;
311	int retval = VMEXIT_CONTINUE;
312
313	newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
314
315        return (retval);
316}
317
318static int
319vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
320{
321	int newcpu;
322	int retval = VMEXIT_CONTINUE;
323
324	newcpu = spinup_ap(ctx, *pvcpu,
325			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
326
327	return (retval);
328}
329
330static int
331vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
332{
333
334	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
335	fprintf(stderr, "\treason\t\tVMX\n");
336	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
337	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
338	fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
339	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
340	fprintf(stderr, "\tqualification\t0x%016lx\n",
341	    vmexit->u.vmx.exit_qualification);
342
343	return (VMEXIT_ABORT);
344}
345
346static int
347vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
348{
349
350	stats.vmexit_bogus++;
351
352	return (VMEXIT_RESTART);
353}
354
355static int
356vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
357{
358
359	stats.vmexit_hlt++;
360
361	/*
362	 * Just continue execution with the next instruction. We use
363	 * the HLT VM exit as a way to be friendly with the host
364	 * scheduler.
365	 */
366	return (VMEXIT_CONTINUE);
367}
368
369static int
370vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
371{
372
373	stats.vmexit_pause++;
374
375	return (VMEXIT_CONTINUE);
376}
377
378static int
379vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
380{
381
382	stats.vmexit_mtrap++;
383
384	return (VMEXIT_RESTART);
385}
386
387static int
388vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
389{
390	int err;
391	stats.vmexit_inst_emul++;
392
393	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
394			  &vmexit->u.inst_emul.vie);
395
396	if (err) {
397		if (err == EINVAL) {
398			fprintf(stderr,
399			    "Failed to emulate instruction at 0x%lx\n",
400			    vmexit->rip);
401		} else if (err == ESRCH) {
402			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
403			    vmexit->u.inst_emul.gpa);
404		}
405
406		return (VMEXIT_ABORT);
407	}
408
409	return (VMEXIT_CONTINUE);
410}
411
412static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
413	[VM_EXITCODE_INOUT]  = vmexit_inout,
414	[VM_EXITCODE_VMX]    = vmexit_vmx,
415	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
416	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
417	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
418	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
419	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
420	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
421};
422
423static void
424vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
425{
426	cpuset_t mask;
427	int error, rc, prevcpu;
428	enum vm_exitcode exitcode;
429
430	if (pincpu >= 0) {
431		CPU_ZERO(&mask);
432		CPU_SET(pincpu + vcpu, &mask);
433		error = pthread_setaffinity_np(pthread_self(),
434					       sizeof(mask), &mask);
435		assert(error == 0);
436	}
437
438	while (1) {
439		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
440		if (error != 0) {
441			/*
442			 * It is possible that 'vmmctl' or some other process
443			 * has transitioned the vcpu to CANNOT_RUN state right
444			 * before we tried to transition it to RUNNING.
445			 *
446			 * This is expected to be temporary so just retry.
447			 */
448			if (errno == EBUSY)
449				continue;
450			else
451				break;
452		}
453
454		prevcpu = vcpu;
455
456		exitcode = vmexit[vcpu].exitcode;
457		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
458			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
459			    exitcode);
460			exit(1);
461		}
462
463                rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
464
465		switch (rc) {
466		case VMEXIT_CONTINUE:
467                        rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
468			break;
469		case VMEXIT_RESTART:
470                        rip = vmexit[vcpu].rip;
471			break;
472		case VMEXIT_RESET:
473			exit(0);
474		default:
475			exit(1);
476		}
477	}
478	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
479}
480
481static int
482num_vcpus_allowed(struct vmctx *ctx)
483{
484	int tmp, error;
485
486	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
487
488	/*
489	 * The guest is allowed to spinup more than one processor only if the
490	 * UNRESTRICTED_GUEST capability is available.
491	 */
492	if (error == 0)
493		return (VM_MAXCPU);
494	else
495		return (1);
496}
497
498void
499fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
500{
501	int err, tmp;
502
503	if (fbsdrun_vmexit_on_hlt()) {
504		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
505		if (err < 0) {
506			fprintf(stderr, "VM exit on HLT not supported\n");
507			exit(1);
508		}
509		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
510		if (cpu == BSP)
511			handler[VM_EXITCODE_HLT] = vmexit_hlt;
512	}
513
514        if (fbsdrun_vmexit_on_pause()) {
515		/*
516		 * pause exit support required for this mode
517		 */
518		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
519		if (err < 0) {
520			fprintf(stderr,
521			    "SMP mux requested, no pause support\n");
522			exit(1);
523		}
524		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
525		if (cpu == BSP)
526			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
527        }
528
529	if (fbsdrun_disable_x2apic())
530		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
531	else
532		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
533
534	if (err) {
535		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
536		exit(1);
537	}
538
539	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
540}
541
542int
543main(int argc, char *argv[])
544{
545	int c, error, gdb_port, err, bvmcons;
546	int max_vcpus;
547	struct vmctx *ctx;
548	uint64_t rip;
549	size_t memsize;
550
551	bvmcons = 0;
552	progname = basename(argv[0]);
553	gdb_port = 0;
554	guest_ncpus = 1;
555	memsize = 256 * MB;
556
557	while ((c = getopt(argc, argv, "abehAHIPWp:g:c:s:S:m:l:")) != -1) {
558		switch (c) {
559		case 'a':
560			disable_x2apic = 1;
561			break;
562		case 'A':
563			acpi = 1;
564			break;
565		case 'b':
566			bvmcons = 1;
567			break;
568		case 'p':
569			pincpu = atoi(optarg);
570			break;
571                case 'c':
572			guest_ncpus = atoi(optarg);
573			break;
574		case 'g':
575			gdb_port = atoi(optarg);
576			break;
577		case 'l':
578			if (lpc_device_parse(optarg) != 0) {
579				errx(EX_USAGE, "invalid lpc device "
580				    "configuration '%s'", optarg);
581			}
582			break;
583		case 's':
584			if (pci_parse_slot(optarg, 0) != 0)
585				exit(1);
586			else
587				break;
588		case 'S':
589			if (pci_parse_slot(optarg, 1) != 0)
590				exit(1);
591			else
592				break;
593                case 'm':
594			error = vm_parse_memsize(optarg, &memsize);
595			if (error)
596				errx(EX_USAGE, "invalid memsize '%s'", optarg);
597			break;
598		case 'H':
599			guest_vmexit_on_hlt = 1;
600			break;
601		case 'I':
602			/*
603			 * The "-I" option was used to add an ioapic to the
604			 * virtual machine.
605			 *
606			 * An ioapic is now provided unconditionally for each
607			 * virtual machine and this option is now deprecated.
608			 */
609			break;
610		case 'P':
611			guest_vmexit_on_pause = 1;
612			break;
613		case 'e':
614			strictio = 1;
615			break;
616		case 'W':
617			virtio_msix = 0;
618			break;
619		case 'h':
620			usage(0);
621		default:
622			usage(1);
623		}
624	}
625	argc -= optind;
626	argv += optind;
627
628	if (argc != 1)
629		usage(1);
630
631	vmname = argv[0];
632
633	ctx = vm_open(vmname);
634	if (ctx == NULL) {
635		perror("vm_open");
636		exit(1);
637	}
638
639	max_vcpus = num_vcpus_allowed(ctx);
640	if (guest_ncpus > max_vcpus) {
641		fprintf(stderr, "%d vCPUs requested but only %d available\n",
642			guest_ncpus, max_vcpus);
643		exit(1);
644	}
645
646	fbsdrun_set_capabilities(ctx, BSP);
647
648	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
649	if (err) {
650		fprintf(stderr, "Unable to setup memory (%d)\n", err);
651		exit(1);
652	}
653
654	init_mem();
655	init_inout();
656	legacy_irq_init();
657
658	rtc_init(ctx);
659
660	/*
661	 * Exit if a device emulation finds an error in it's initilization
662	 */
663	if (init_pci(ctx) != 0)
664		exit(1);
665
666	ioapic_init(0);
667
668	if (gdb_port != 0)
669		init_dbgport(gdb_port);
670
671	if (bvmcons)
672		init_bvmcons();
673
674	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
675	assert(error == 0);
676
677	/*
678	 * build the guest tables, MP etc.
679	 */
680	mptable_build(ctx, guest_ncpus);
681
682	if (acpi) {
683		error = acpi_build(ctx, guest_ncpus);
684		assert(error == 0);
685	}
686
687	/*
688	 * Change the proc title to include the VM name.
689	 */
690	setproctitle("%s", vmname);
691
692	/*
693	 * Add CPU 0
694	 */
695	fbsdrun_addcpu(ctx, BSP, rip);
696
697	/*
698	 * Head off to the main event dispatch loop
699	 */
700	mevent_dispatch();
701
702	exit(1);
703}
704